diff options
Diffstat (limited to 'SingleSource/UnitTests/Vector/AVX512F/fma.c')
| -rw-r--r-- | SingleSource/UnitTests/Vector/AVX512F/fma.c | 827 |
1 files changed, 827 insertions, 0 deletions
diff --git a/SingleSource/UnitTests/Vector/AVX512F/fma.c b/SingleSource/UnitTests/Vector/AVX512F/fma.c new file mode 100644 index 00000000..73ef4b6d --- /dev/null +++ b/SingleSource/UnitTests/Vector/AVX512F/fma.c @@ -0,0 +1,827 @@ +/* + * Test fma(fmadd, fmsub, fnmadd, fnmsub) instructions. + * Here we check for _mm512_[mask|mask3]_f[madd|msub|nmadd|nmsub]_[round] + * intrinsics. + */ + +#include "m512_test_util.h" +#include <stdio.h> +#include <string.h> + +int verbose = 0; + +__m512i i1; +__m512i i2; +__m512i i3; +__m512i i4; +__m512i i5; + +__m512 f1; +__m512 f2; +__m512 f3; +__m512 f4; +__m512 f5; + +__m512d d1; +__m512d d2; +__m512d d3; +__m512d d4; +__m512d d5; + +typedef enum { FMA_233, FMA_132, FMA_231, FMA_213, FMA_23c1 } Fma_order; + +volatile int vol = 0; /* To prevent optimizations */ + +void NOINLINE init() { + int i; + V512 *pi1 = (V512 *)&i1; + V512 *pi2 = (V512 *)&i2; + V512 *pi3 = (V512 *)&i3; + V512 *pf1 = (V512 *)&f1; + V512 *pf2 = (V512 *)&f2; + V512 *pf3 = (V512 *)&f3; + V512 *pd1 = (V512 *)&d1; + V512 *pd2 = (V512 *)&d2; + V512 *pd3 = (V512 *)&d3; + + for (i = 0; i < 16; i++) { + pi1->s32[i] = 17 + ((i & 1) ? 1 : -1) * i + vol; + pf1->f32[i] = pi1->s32[i]; + + pi2->s32[i] = 100 + ((i & 3) == 3 ? 1 : -1) * i + vol; + pf2->f32[i] = -pi2->s32[i]; + + pi3->s32[i] = 400 + ((i & 1) ? -1 : 1) * i + vol; + pf3->f32[i] = pi3->s32[i]; + } + + for (i = 0; i < 8; i++) { + pd1->f64[i] = pi1->s32[i]; + pd2->f64[i] = pi2->s32[i]; + pd3->f64[i] = -pi3->s32[i]; + } +} + +void NOINLINE check_equal32(void *vgot, void *vexpected, void *vexpected_orig, + int mask, char *banner) { + int i; + V512 *got = (V512 *)vgot; + V512 *expected = (V512 *)vexpected; + V512 *orig = (V512 *)vexpected_orig; + + for (i = 0; i < 16; i++) { + int ans = (mask & (1 << i)) ? expected->s32[i] : orig->s32[i]; + if (got->s32[i] != ans) { + printf("ERROR: %s failed -- 0x%0.8x != 0x%0.8x at element [%d]\n", + banner ? banner : "", got->s32[i], ans, i); + n_errs++; + break; + } + } +} + +void NOINLINE check_equal64(void *vgot, void *vexpected, void *vexpected_orig, + int mask, char *banner) { + int i; + V512 *got = (V512 *)vgot; + V512 *expected = (V512 *)vexpected; + V512 *orig = (V512 *)vexpected_orig; + + for (i = 0; i < 8; i++) { + __int64 ans = (mask & (1 << i)) ? expected->s64[i] : orig->s64[i]; + if (got->s64[i] != ans) { + printf("ERROR: %s failed -- %0.16" PRIx64 " != %0.16" PRIx64 + " at element [%d]\n", + banner ? banner : "", got->s64[i], ans, i); + n_errs++; + break; + } + } +} + +void NOINLINE emulate_fmadd_ps(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 16; i++) { + + if (((1 << i) & mask) == 0) { + result->u32[i] = v1->u32[i]; + continue; + } + + switch (order) { + case FMA_233: + result->f32[i] = + v2->f32[i] * v3->f32[(i & ~0x3) + 1] + v3->f32[(i & ~0x3)]; + break; + + case FMA_132: + result->f32[i] = v1->f32[i] * v3->f32[i] + v2->f32[i]; + break; + + case FMA_231: + result->f32[i] = v2->f32[i] * v3->f32[i] + v1->f32[i]; + break; + + case FMA_213: + result->f32[i] = v2->f32[i] * v1->f32[i] + v3->f32[i]; + break; + + case FMA_23c1: + result->f32[i] = (v2->f32[i] * v3->f32[i]) + 1.0f; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE emulate_fmsub_ps(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 16; i++) { + + if (((1 << i) & mask) == 0) { + result->u32[i] = v1->u32[i]; + continue; + } + + switch (order) { + case FMA_233: + result->f32[i] = + v2->f32[i] * v3->f32[(i & ~0x3) + 1] - v3->f32[(i & ~0x3)]; + break; + + case FMA_132: + result->f32[i] = v1->f32[i] * v3->f32[i] - v2->f32[i]; + break; + + case FMA_231: + result->f32[i] = v2->f32[i] * v3->f32[i] - v1->f32[i]; + break; + + case FMA_213: + result->f32[i] = v2->f32[i] * v1->f32[i] - v3->f32[i]; + break; + + case FMA_23c1: + result->f32[i] = (v2->f32[i] * v3->f32[i]) - 1.0f; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE emulate_fnmadd_ps(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 16; i++) { + + if (((1 << i) & mask) == 0) { + result->u32[i] = v1->u32[i]; + continue; + } + + switch (order) { + case FMA_233: + result->f32[i] = + -(v2->f32[i] * v3->f32[(i & ~0x3) + 1]) + v3->f32[(i & ~0x3)]; + break; + + case FMA_132: + result->f32[i] = -(v1->f32[i] * v3->f32[i]) + v2->f32[i]; + break; + + case FMA_231: + result->f32[i] = -(v2->f32[i] * v3->f32[i]) + v1->f32[i]; + break; + + case FMA_213: + result->f32[i] = -(v2->f32[i] * v1->f32[i]) + v3->f32[i]; + break; + + case FMA_23c1: + result->f32[i] = -(v2->f32[i] * v3->f32[i]) + 1.0f; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE emulate_fnmsub_ps(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 16; i++) { + + if (((1 << i) & mask) == 0) { + result->u32[i] = v1->u32[i]; + continue; + } + + switch (order) { + case FMA_233: + result->f32[i] = + -(v2->f32[i] * v3->f32[(i & ~0x3) + 1]) - v3->f32[(i & ~0x3)]; + break; + + case FMA_132: + result->f32[i] = -(v1->f32[i] * v3->f32[i]) - v2->f32[i]; + break; + + case FMA_231: + result->f32[i] = -(v2->f32[i] * v3->f32[i]) - v1->f32[i]; + break; + + case FMA_213: + result->f32[i] = -(v2->f32[i] * v1->f32[i]) - v3->f32[i]; + break; + + case FMA_23c1: + result->f32[i] = -(v2->f32[i] * v3->f32[i]) - 1.0f; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE emulate_fmadd_pi(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 16; i++) { + + if (((1 << i) & mask) == 0) { + result->u32[i] = v1->u32[i]; + continue; + } + + switch (order) { + case FMA_233: + result->s32[i] = + v2->s32[i] * v3->s32[(i & ~0x3) + 1] + v3->s32[(i & ~0x3)]; + break; + + case FMA_132: + result->s32[i] = v1->s32[i] * v3->s32[i] + v2->s32[i]; + break; + + case FMA_231: + result->s32[i] = v2->s32[i] * v3->s32[i] + v1->s32[i]; + break; + + case FMA_213: + result->s32[i] = v2->s32[i] * v1->s32[i] + v3->s32[i]; + break; + + case FMA_23c1: + result->s32[i] = v2->s32[i] * v3->s32[i] + 1; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE emulate_fmadd_pd(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 8; i++) { + + if (((1 << i) & mask) == 0) { + result->u64[i] = v1->u64[i]; + continue; + } + + switch (order) { + case FMA_233: + result->f64[i] = + v2->f64[i] * v3->f64[(i & ~0x3) + 1] + v3->f64[(i & ~0x3)]; + break; + + case FMA_132: + result->f64[i] = v1->f64[i] * v3->f64[i] + v2->f64[i]; + break; + + case FMA_231: + result->f64[i] = v2->f64[i] * v3->f64[i] + v1->f64[i]; + break; + + case FMA_213: + result->f64[i] = v2->f64[i] * v1->f64[i] + v3->f64[i]; + break; + + case FMA_23c1: + result->f64[i] = v2->f64[i] * v3->f64[i] + 1.0; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE emulate_fmsub_pd(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 8; i++) { + + if (((1 << i) & mask) == 0) { + result->u64[i] = v1->u64[i]; + continue; + } + + switch (order) { + case FMA_233: + result->f64[i] = + v2->f64[i] * v3->f64[(i & ~0x3) + 1] - v3->f64[(i & ~0x3)]; + break; + + case FMA_132: + result->f64[i] = v1->f64[i] * v3->f64[i] - v2->f64[i]; + break; + + case FMA_231: + result->f64[i] = v2->f64[i] * v3->f64[i] - v1->f64[i]; + break; + + case FMA_213: + result->f64[i] = v2->f64[i] * v1->f64[i] - v3->f64[i]; + break; + + case FMA_23c1: + result->f64[i] = v2->f64[i] * v3->f64[i] - 1.0; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE emulate_fnmadd_pd(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 8; i++) { + + if (((1 << i) & mask) == 0) { + result->u64[i] = v1->u64[i]; + continue; + } + + switch (order) { + case FMA_233: + result->f64[i] = + -(v2->f64[i] * v3->f64[(i & ~0x3) + 1]) + v3->f64[(i & ~0x3)]; + break; + + case FMA_132: + result->f64[i] = -(v1->f64[i] * v3->f64[i]) + v2->f64[i]; + break; + + case FMA_231: + result->f64[i] = -(v2->f64[i] * v3->f64[i]) + v1->f64[i]; + break; + + case FMA_213: + result->f64[i] = -(v2->f64[i] * v1->f64[i]) + v3->f64[i]; + break; + + case FMA_23c1: + result->f64[i] = -(v2->f64[i] * v3->f64[i]) + 1.0; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE emulate_fnmsub_pd(void *presult, const void *p1, int mask, + const void *p2, const void *p3, + Fma_order order) { + int i; + V512 *result = (V512 *)presult; + V512 *v1 = (V512 *)p1; + V512 *v2 = (V512 *)p2; + V512 *v3 = (V512 *)p3; + + for (i = 0; i < 8; i++) { + + if (((1 << i) & mask) == 0) { + result->u64[i] = v1->u64[i]; + continue; + } + + switch (order) { + case FMA_233: + result->f64[i] = + -(v2->f64[i] * v3->f64[(i & ~0x3) + 1]) - v3->f64[(i & ~0x3)]; + break; + + case FMA_132: + result->f64[i] = -(v1->f64[i] * v3->f64[i]) - v2->f64[i]; + break; + + case FMA_231: + result->f64[i] = -(v2->f64[i] * v3->f64[i]) - v1->f64[i]; + break; + + case FMA_213: + result->f64[i] = -(v2->f64[i] * v1->f64[i]) - v3->f64[i]; + break; + + case FMA_23c1: + result->f64[i] = -(v2->f64[i] * v3->f64[i]) - 1.0; + break; + + default: + printf("ERROR -- bad fma order %d\n", (int)order); + n_errs++; + return; + } + } +} + +void NOINLINE do_fmadd_ps() { + f4 = _mm512_fmadd_ps(f1, f2, f3); + emulate_fmadd_ps(&f5, &f1, 0xffff, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, (void *)0, 0xffff, "_mm512_fmadd_ps"); + + f4 = _mm512_mask_fmadd_ps(f1, 0x79fa, f2, f3); + emulate_fmadd_ps(&f5, &f1, 0x79fa, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, &f1, 0x79fa, "_mm512_mask_fmadd_ps"); + + f4 = _mm512_mask3_fmadd_ps(f1, f2, f3, 0x563a); + emulate_fmadd_ps(&f5, &f3, 0x563a, &f1, &f2, FMA_231); + check_equal32(&f4, &f5, &f3, 0x563a, "_mm512_mask3_fmadd_ps"); + + /* + * Employ rounding modes. + * Our FP inputs are all integer values, so there's no need for any + * special emulation routine. + */ + + f4 = _mm512_fmadd_round_ps(f1, f2, f3, + _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); + emulate_fmadd_ps(&f5, &f1, 0xffff, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, (void *)0, 0xffff, "_mm512_fmadd_round_ps"); + + f4 = _mm512_mask_fmadd_round_ps(f1, 0x79fa, f2, f3, + _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC); + emulate_fmadd_ps(&f5, &f1, 0x79fa, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, &f1, 0x79fa, "_mm512_mask_fmadd_round_ps"); + + f4 = _mm512_mask3_fmadd_round_ps(f1, f2, f3, 0x563a, + _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC); + emulate_fmadd_ps(&f5, &f3, 0x563a, &f1, &f2, FMA_231); + check_equal32(&f4, &f5, &f3, 0x563a, "_mm512_mask3_fmadd_round_ps"); +} + +void NOINLINE do_fnmsub_ps() { + f4 = _mm512_fnmsub_ps(f1, f2, f3); + emulate_fnmsub_ps(&f5, &f1, 0xffff, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, (void *)0, 0xffff, "_mm512_fnmsub_ps"); + + f4 = _mm512_mask_fnmsub_ps(f1, 0x79fa, f2, f3); + emulate_fnmsub_ps(&f5, &f1, 0x79fa, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, &f1, 0x79fa, "_mm512_mask_fnmsub_ps"); + + f4 = _mm512_mask3_fnmsub_ps(f1, f2, f3, 0x563a); + emulate_fnmsub_ps(&f5, &f3, 0x563a, &f1, &f2, FMA_231); + check_equal32(&f4, &f5, &f3, 0x563a, "_mm512_mask3_fnmsub_ps"); + + /* + * Employ rounding modes. + * Our FP inputs are all integer values, so there's no need for any + * special emulation routine. + */ + + f4 = _mm512_fnmsub_round_ps(f1, f2, f3, + _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); + emulate_fnmsub_ps(&f5, &f1, 0xffff, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, (void *)0, 0xffff, "_mm512_fnmsub_round_ps"); + + f4 = _mm512_mask_fnmsub_round_ps(f1, 0x79fa, f2, f3, + _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC); + emulate_fnmsub_ps(&f5, &f1, 0x79fa, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, &f1, 0x79fa, "_mm512_mask_fnmsub_round_ps"); + + f4 = _mm512_mask3_fnmsub_round_ps(f1, f2, f3, 0x563a, + _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC); + emulate_fnmsub_ps(&f5, &f3, 0x563a, &f1, &f2, FMA_231); + check_equal32(&f4, &f5, &f3, 0x563a, "_mm512_mask3_fnmsub_round_ps"); +} + +void NOINLINE do_fmadd_pd() { + d4 = _mm512_fmadd_pd(d1, d2, d3); + emulate_fmadd_pd(&d5, &d1, 0xffff, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, (void *)0, 0xffff, "_mm512_fmadd_pd"); + + d4 = _mm512_mask_fmadd_pd(d1, 0xfa, d2, d3); + emulate_fmadd_pd(&d5, &d1, 0xfa, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, &d1, 0xfa, "_mm512_mask_fmadd_pd"); + + d4 = _mm512_mask3_fmadd_pd(d1, d2, d3, 0x56); + emulate_fmadd_pd(&d5, &d3, 0x56, &d1, &d2, FMA_231); + check_equal64(&d4, &d5, &d3, 0x56, "_mm512_mask3_fmadd_pd"); + + /* + * Employ rounding modes. + * Our FP inputs are all integer values, so there's no need for any + * special emulation routine. + */ + + d4 = _mm512_fmadd_round_pd(d1, d2, d3, + _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); + emulate_fmadd_pd(&d5, &d1, 0xffff, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, (void *)0, 0xffff, "_mm512_fmadd_round_pd"); + + d4 = _mm512_mask_fmadd_round_pd(d1, 0x79, d2, d3, + _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC); + emulate_fmadd_pd(&d5, &d1, 0x79, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, &d1, 0x79, "_mm512_mask_fmadd_round_pd"); + + d4 = _mm512_mask3_fmadd_round_pd(d1, d2, d3, 0x63, + _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC); + emulate_fmadd_pd(&d5, &d3, 0x63, &d1, &d2, FMA_231); + check_equal64(&d4, &d5, &d3, 0x63, "_mm512_mask3_fmadd_round_pd"); +} + +void NOINLINE do_fnmsub_pd() { + d4 = _mm512_fnmsub_pd(d1, d2, d3); + emulate_fnmsub_pd(&d5, &d1, 0xffff, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, (void *)0, 0xffff, "_mm512_fnmsub_pd"); + + d4 = _mm512_mask_fnmsub_pd(d1, 0xfa, d2, d3); + emulate_fnmsub_pd(&d5, &d1, 0xfa, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, &d1, 0xfa, "_mm512_mask_fnmsub_pd"); + + d4 = _mm512_mask3_fnmsub_pd(d1, d2, d3, 0x56); + emulate_fnmsub_pd(&d5, &d3, 0x56, &d1, &d2, FMA_231); + check_equal64(&d4, &d5, &d3, 0x56, "_mm512_mask3_fnmsub_pd"); + + /* + * Employ rounding modes. + * Our FP inputs are all integer values, so there's no need for any + * special emulation routine. + */ + + d4 = _mm512_fnmsub_round_pd(d1, d2, d3, + _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); + emulate_fnmsub_pd(&d5, &d1, 0xffff, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, (void *)0, 0xffff, "_mm512_fnmsub_round_pd"); + + d4 = _mm512_mask_fnmsub_round_pd(d1, 0x79, d2, d3, + _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC); + emulate_fnmsub_pd(&d5, &d1, 0x79, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, &d1, 0x79, "_mm512_mask_fnmsub_round_pd"); + + d4 = _mm512_mask3_fnmsub_round_pd(d1, d2, d3, 0x63, + _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC); + emulate_fnmsub_pd(&d5, &d3, 0x63, &d1, &d2, FMA_231); + check_equal64(&d4, &d5, &d3, 0x63, "_mm512_mask3_fnmsub_round_pd"); +} + +void NOINLINE do_fmsub_ps() { + f4 = _mm512_fmsub_ps(f1, f2, f3); + emulate_fmsub_ps(&f5, &f1, 0xffff, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, (void *)0, 0xffff, "_mm512_fmsub_ps"); + + f4 = _mm512_mask_fmsub_ps(f1, 0x79fa, f2, f3); + emulate_fmsub_ps(&f5, &f1, 0x79fa, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, &f1, 0x79fa, "_mm512_mask_fmsub_ps"); + + f4 = _mm512_mask3_fmsub_ps(f1, f2, f3, 0x563a); + emulate_fmsub_ps(&f5, &f3, 0x563a, &f1, &f2, FMA_231); + check_equal32(&f4, &f5, &f3, 0x563a, "_mm512_mask3_fmsub_ps"); + + /* + * Employ rounding modes. + * Our FP inputs are all integer values, so there's no need for any + * special emulation routine. + */ + + f4 = _mm512_fmsub_round_ps(f1, f2, f3, + _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); + emulate_fmsub_ps(&f5, &f1, 0xffff, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, (void *)0, 0xffff, "_mm512_fmsub_round_ps"); + + f4 = _mm512_mask_fmsub_round_ps(f1, 0x79fa, f2, f3, + _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC); + emulate_fmsub_ps(&f5, &f1, 0x79fa, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, &f1, 0x79fa, "_mm512_mask_fmsub_round_ps"); + + f4 = _mm512_mask3_fmsub_round_ps(f1, f2, f3, 0x563a, + _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC); + emulate_fmsub_ps(&f5, &f3, 0x563a, &f1, &f2, FMA_231); + check_equal32(&f4, &f5, &f3, 0x563a, "_mm512_mask3_fmsub_round_ps"); +} + +void NOINLINE do_fnmadd_ps() { + f4 = _mm512_fnmadd_ps(f1, f2, f3); + emulate_fnmadd_ps(&f5, &f1, 0xffff, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, (void *)0, 0xffff, "_mm512_fnmadd_ps"); + + f4 = _mm512_mask_fnmadd_ps(f1, 0x79fa, f2, f3); + emulate_fnmadd_ps(&f5, &f1, 0x79fa, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, &f1, 0x79fa, "_mm512_mask_fnmadd_ps"); + + f4 = _mm512_mask3_fnmadd_ps(f1, f2, f3, 0x563a); + emulate_fnmadd_ps(&f5, &f3, 0x563a, &f1, &f2, FMA_231); + check_equal32(&f4, &f5, &f3, 0x563a, "_mm512_mask3_fnmadd_ps"); + + /* + * Employ rounding modes. + * Our FP inputs are all integer values, so there's no need for any + * special emulation routine. + */ + + f4 = _mm512_fnmadd_round_ps(f1, f2, f3, + _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); + emulate_fnmadd_ps(&f5, &f1, 0xffff, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, (void *)0, 0xffff, "_mm512_fnmadd_round_ps"); + + f4 = _mm512_mask_fnmadd_round_ps(f1, 0x79fa, f2, f3, + _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC); + emulate_fnmadd_ps(&f5, &f1, 0x79fa, &f3, &f2, FMA_132); + check_equal32(&f4, &f5, &f1, 0x79fa, "_mm512_mask_fnmadd_round_ps"); + + f4 = _mm512_mask3_fnmadd_round_ps(f1, f2, f3, 0x563a, + _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC); + emulate_fnmadd_ps(&f5, &f3, 0x563a, &f1, &f2, FMA_231); + check_equal32(&f4, &f5, &f3, 0x563a, "_mm512_mask3_fnmadd_round_ps"); +} + +void NOINLINE do_fmsub_pd() { + d4 = _mm512_fmsub_pd(d1, d2, d3); + emulate_fmsub_pd(&d5, &d1, 0xffff, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, (void *)0, 0xffff, "_mm512_fmsub_pd"); + + d4 = _mm512_mask_fmsub_pd(d1, 0xfa, d2, d3); + emulate_fmsub_pd(&d5, &d1, 0xfa, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, &d1, 0xfa, "_mm512_mask_fmsub_pd"); + + d4 = _mm512_mask3_fmsub_pd(d1, d2, d3, 0x56); + emulate_fmsub_pd(&d5, &d3, 0x56, &d1, &d2, FMA_231); + check_equal64(&d4, &d5, &d3, 0x56, "_mm512_mask3_fmsub_pd"); + + /* + * Employ rounding modes. + * Our FP inputs are all integer values, so there's no need for any + * special emulation routine. + */ + + d4 = _mm512_fmsub_round_pd(d1, d2, d3, + _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); + emulate_fmsub_pd(&d5, &d1, 0xffff, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, (void *)0, 0xffff, "_mm512_fmsub_round_pd"); + + d4 = _mm512_mask_fmsub_round_pd(d1, 0x79, d2, d3, + _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC); + emulate_fmsub_pd(&d5, &d1, 0x79, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, &d1, 0x79, "_mm512_mask_fmsub_round_pd"); + + d4 = _mm512_mask3_fmsub_round_pd(d1, d2, d3, 0x63, + _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC); + emulate_fmsub_pd(&d5, &d3, 0x63, &d1, &d2, FMA_231); + check_equal64(&d4, &d5, &d3, 0x63, "_mm512_mask3_fmsub_round_pd"); +} + +void NOINLINE do_fnmadd_pd() { + d4 = _mm512_fnmadd_pd(d1, d2, d3); + emulate_fnmadd_pd(&d5, &d1, 0xffff, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, (void *)0, 0xffff, "_mm512_fnmadd_pd"); + + d4 = _mm512_mask_fnmadd_pd(d1, 0xfa, d2, d3); + emulate_fnmadd_pd(&d5, &d1, 0xfa, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, &d1, 0xfa, "_mm512_mask_fnmadd_pd"); + + d4 = _mm512_mask3_fnmadd_pd(d1, d2, d3, 0x56); + emulate_fnmadd_pd(&d5, &d3, 0x56, &d1, &d2, FMA_231); + check_equal64(&d4, &d5, &d3, 0x56, "_mm512_mask3_fnmadd_pd"); + + /* + * Employ rounding modes. + * Our FP inputs are all integer values, so there's no need for any + * special emulation routine. + */ + + d4 = _mm512_fnmadd_round_pd(d1, d2, d3, + _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC); + emulate_fnmadd_pd(&d5, &d1, 0xffff, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, (void *)0, 0xffff, "_mm512_fnmadd_round_pd"); + + d4 = _mm512_mask_fnmadd_round_pd(d1, 0x79, d2, d3, + _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC); + emulate_fnmadd_pd(&d5, &d1, 0x79, &d3, &d2, FMA_132); + check_equal64(&d4, &d5, &d1, 0x79, "_mm512_mask_fnmadd_round_pd"); + + d4 = _mm512_mask3_fnmadd_round_pd(d1, d2, d3, 0x63, + _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC); + emulate_fnmadd_pd(&d5, &d3, 0x63, &d1, &d2, FMA_231); + check_equal64(&d4, &d5, &d3, 0x63, "_mm512_mask3_fnmadd_round_pd"); +} + +int main(int argc, char *argv[]) { + if (argc > 1 && argv[1][0] == '-' && argv[1][1] == 'v' && + argv[1][2] == '\0') { + verbose = 1; + } + + init(); + + do_fmadd_ps(); + + do_fmadd_pd(); + + do_fnmsub_ps(); + + do_fnmsub_pd(); + + do_fmsub_ps(); + + do_fmsub_pd(); + + do_fnmadd_ps(); + + do_fnmadd_pd(); + + if (n_errs != 0) { + printf("FAILED\n"); + return 1; + } + + printf("PASSED\n"); + return 0; +} |